Speculative query executions can be a particularly valuable technique for addressing a number of Cassandra database issues — from faulty, slow, or unresponsive nodes, to network interruptions. Using speculative query execution allows a client to make database requests from multiple endpoints at the same time and have the requests compete to see which provides the quickest response. While setting up this race between requests does determine which node is most performant, performance isn’t usually the goal of a speculative query execution. Instead, the purpose is to make sure that queries receive successful server responses (streamlined execution time can certainly be a happy byproduct, however).
Important note before we go any further: speculative queries require the use of CPU and network resources, so it’s important to remember that the reliability or performance improvements they may yield aren’t without a cost.